面向对象的高空间分辨率遥感土地覆盖分类
Object-Oriented Land Cover Classification Using High Spatial Resolution Remote Sensing

高空间分辨率遥感影像数据已成为快速获取地表信息的主要数据源。面向对象的分类方法，充分利用了高空间分辨率遥感影像丰富的光谱、形状、纹理等特征，通过影像多尺度分割、多特征分析与提取、样本采集、监督分类等技术，得到较高精度的分类结果。本文选取浙江省舟山岛2016年第一季度的GF-1高空间分辨率遥感影像数据，结合面向对象技术和C5.0决策树算法对研究区进行土地覆盖分类研究。结果表明，分类总体精度达到91%，Kappa系数达到0.87，具有可靠的分类精度，能够实现土地覆盖精准、快速的自动识别分类。
High spatial resolution remote sensing images have become the main data sources to achieve the extraction of ground information. Object-oriented classification method makes use of the features of shape, texture, landscape to realize high precision classification results by the technologies of images segmentation, multiple feature analysis and extraction, sample selection, supervised classification. According to the GF-1 high spatial resolution remote sensing images, this paper combined object-oriented method with C5.0 decision tree algorithm to study land cover classification of Zhoushan Island, Zhejiang Province in 2016. The experimental results show that the total accuracy of classification result was 91%, and the Kappa coefficient was 0.87. The classification result had credible precision. The method was a high precision and rapid automatic identification classification method.